Defining familial longevity and developing a familial longevity score for unbiased epigenetic studies in a birth cohort

Abstract

Aim: Longevity accumulating in families has genetic and epigenetic components. To study early and unbiased epigenetic predictors of longevity prospectively, a birth cohort would be ideal. However, the original family longevity selection score (FLoSS) focuses on populations of elderly only.

Methods: In the German birth cohort KUNO-Kids we assessed when information for such scores may be best collected and how to calculate an adapted FLoSS.

Results: A total of 551 families contributed to adapted FLoSS, with a mean score of -0.15 (SD 2.33). Adapted FLoSS ≥7 as a marker of exceptional longevity occurred in 3.3% of families, comparable to original FLoSS in elderly.

Conclusion: An adapted FLoSS from data collectable postnatally may be a feasible tool to study unbiased epigenetic predictors for longevity.

Tweetable Abstract

In the German birth cohort KUNO-Kids we assessed if and how a family longevity selection score may best be calculated to study unbiased epigenetic predictors for longevity in the future.

Plain language summary

Article highlights.

• Longevity accumulates in families and genetic as well as epigenetic markers for longevity are overrepresented in some families.

• To study epigenetic effects causally involved in longevity without a bias, it has to be studied early in life and birth cohorts may be a good chance to do so.

• However, existing longevity scores have not been developed for that task and it is unclear if they could work in the setting of birth cohort or if the information on previous generations needed to calculate such scores can be collected in birth cohorts.

• We used the German KUNO Kids Health study, which recruited more than 3000 children since 2015 to test if and when data on longevity can be collected from parents of a newborn and found that parents can give sufficient data on longevity in their families right after birth.

• We used these data to built an adapted family longevity selection score (FLoSS).

• We compared the performance of our adapted FLoSS in the KUNO Kids Study to the initial score in classical settings of study populations of old people and found that it lead to very similar distributions of families with aggravated longevity.

• We compared results of the adapted Floss in detecting families with aggravated longevity to other scores such as Longevity Relatives Count and found that FloSS performs better in birth cohorts.

• We conclude that using adapted FLoSS to identify families with aggravated longevity at the birth of a child is feasible and a promising opportunity to study early and unbiased epigenetic markers of longevity in the future.

1. Introduction

Longevity seems to have a strong genetic component [1–4]. Familial longevity is one of the most important predictors for survival to an age of 100 years and beyond for the offspring [5]. Longevity accumulates in families [6] and genetic markers for longevity are overrepresented in some families [7,8]. In addition to genetic factors, epigenetic mechanisms have recently been suspected to play an important role in the aging process [3,4,9–12]: Epigenetic mechanisms, such as methylation of DNA, modifications of histones and microRNA may be able to change gene function for a lifetime and even beyond [4], without altering DNA sequence in the process [3,9–12]. From embryonic time to old age [13], influences and behaviors [9] such as stress [10], psychosocial factors, diet [10,14,15] and physical activity [9] imprint on the individual's epigenetic profile and may impact on health in general and ageing in particular [10,16–18].

Simply counting old people in a family is not sufficient for the purpose, as a number of influences not related to exceptional longevity must be accounted for according to generation-specific effects. Thus, the original family longevity selection score (FLoSS) was developed to define families' survival based on a very old index individual and the number of living very old siblings [19]. More recent studies confirmed the applicability of the original FLoSS in a multigenerational approach over two [20] and even three generations of elderly people [6]. Further scores such as the Longevity Relatives Count (LRC) score have also been proposed to calculate longevity in families to study inheritance [6,21].

However, when epigenetic factors influencing longevity are investigated in middle-aged or old individuals, a bias is introduced, as environmental exposures, events and behaviors over lifetime have already influenced the individual's epigenetic signature. If these markers should be discriminated as much as possible from ‘inborn’ epigenetic predictors which may be causative (and probably inherited over generations) as well as changeable in the individual, one would have to study epigenetic signatures in the very young and find a method to predict and score longevity in the child's family. For this purpose, an adaptation of the FLoSS to the situation of birth cohorts would be desirable and needed to have a readily available longevity outcome for birth cohort studies in the absence of information on old sib pairs (usually not available in birth cohorts), on which the original FLoSS is based.

Thus, we used KUNO-Kids Study (KKS) data [22] to test if the original FLoSS could be adapted successfully to assess familial longevity in a birth cohort. Starting with the newborn, we developed questionnaires to assess longevity over multigenerational families and adapted the FLoSS accordingly. We then compared our results to studies in which the original FLoSS was applied to populations of elderly with a similar European genetic background from the general population or populations enriched for longevity to test if our adapted FloSS would yield plausible numbers of long-lived families when compared with these reference populations. Furthermore, we compared the adapted FloSS to the LRC score in our birth cohort KKS.

2. Materials & methods

2.1. KUNO-Kids health study

This study was performed in the KKS, a prospective birth cohort study in Eastern Bavaria, Germany, a region of approximately 2 million inhabitants described in more detail previously [22]. Briefly, more than 3000 newborns and their families have been included in the study since 2015 and assessed for various health outcomes and factors influencing health and disease starting with their mother’s pregnancy. Biomaterial was sampled at birth, including cord blood, stool, urine, skin and buccal swabs. All study participants gave their written informed consent and all study procedures were approved by the medical ethics committee of the university of Regensburg (file number: 14-101-0347 and 19-1646-101).

2.2. Development of longevity questionnaires & timing of data acquisition in a birth cohort setting

We developed three questionnaires to collect data for the analysis of familial longevity, which were handed out at three different time points (within 48 h after birth, 4 weeks after birth and 1 year after birth) to evaluate the best time to acquire the most reliable and complete data (Figure 1 & Supplementary Table S1). In the KKS, an interview with general questions was performed within 48 h after birth at the maternity unit and a first longevity questionnaire was distributed right after the interview containing three items (date of birth [in decades], still alive or date of death) for 12 relatives (four grandparents and eight great-grandparents). The reason of death was inquired and categorized by senility, death by illness or unnatural cause of death (e.g., due to war, accident or suicide).

Figure 1.

Participation at different survey times (A) for the overall KKS and (B) response rates and quality of responses for the subsample of 50 randomly selected families from KKS. (A) shows participation rates of KKS families at the three different survey times. The first questionnaire was answered at the hospital right after birth by the families. For the following surveys, questionnaires were sent to mothers and fathers separately. (B) shows results of the subsample of 50 randomly selected families from KKS. There plausibility and completeness of data was assessed based on a question for grandparents' birthyear which was available at all three timepoints.

KKS: KUNO-Kids Study.

Four weeks after discharge, a further questionnaire including eight items about familial longevity was sent to the families (separately to mother and father) as part of a larger, general KKS 4-weeks questionnaire, including country of birth, exact years of birth and death of grandparents and more detailed questions about illnesses.

Finally, a third questionnaire was sent as part of the regular 1-year KKS questionnaire, to mothers and fathers with five items asking for year of birth and exact dates of death for both grand- and great-grandparents. Apart from that, the content of the questionnaire was identical to the questionnaire after birth. All questionnaires were sent to the same participants.

2.3. Adaptation of the original FLoSS to a birth cohort setting

The above questions had been developed with the purpose to be used in an adaptation of the original FLoSS [19] to the birth cohort setting. The original FLoSS ranks families for longevity and has two elements: (1) an estimated family longevity score built from birth-, gender- and nation-specific cohort survival probabilities and (2) a living sibling bonus [19]. The mathematical formulas of the original FLoSS and the adapted FLoSS as well as their specific differences are shown and compared in Table 1. The information for the original score had been collected with the help of living relatives and consists of the number of siblings, number of living siblings, average-age and an exceptional lifespan. The year of birth, year of death, gender and country of birth are also required to calculate the score. Since the score excluded people under the age of 40, parents of the newborns and people who died before the age of 40 were excluded from the calculation of the original FLoSS score [19].

Table 1.

Modifications to the FLoSS for a birth cohort.

	FLoSS	KKS-FLoSS
Participants	One sibship	More generations of a family
Generations	1	2
Starting point	Old person	Newborn
Sibling	Essential	Not required
Reference cohort	Birthcohort with same birthyear, gender and country	Birthcohort with same birthyear and gender
Dead family members	Deceased siblings	Deceased grand- and great-grandparents
Living family members	Living siblings	Living grand- and great-grandparents
Bonus	Bonus for old living siblings	Bonus for old living grand- and great-grandparents
Formula	$\begin{array}{c}\mathrm{FLoS}{S}_f\\ ={\Sigma }_{\mathrm{dead}\mathrm{family}\mathrm{members}}\left(-\ln \left(p\left(A|C\right)\right)-1\right)\\ +{\Sigma }_{\mathrm{living}\mathrm{family}\mathrm{members}}\left[\right(-ln\left(p\left(A^{*}|C\right)\right)-1)\\ +max\left\{0,-ln\left(p\left(A|C\right)\right)-1\right\}]\end{array}$	$\begin{array}{c}\mathrm{FLoS}{S}_{KKS}\\ ={\Sigma }_{\mathrm{dead}\mathrm{grandparents}}\left(-ln\left(p\left(A|C\right)\right)-1\right)\\ +{\Sigma }_{\mathrm{living}\mathrm{grandparents}}\left[\right(-ln\left(p\left(A^{*}|C\right)\right)-1)\\ +max\left\{0,-ln\left(p\left(A|C\right)\right)-1\right\}]\\ +{\Sigma }_{\mathrm{dead}\mathrm{great}-\mathrm{grandparents}}\left(-ln\left(p\left(A|C\right)\right)-1\right)\\ +{\Sigma }_{\mathrm{living}\mathrm{great}-\mathrm{grandparents}}\left[\right(-ln\left(p\left(A^{*}|C\right)\right)-1)\\ +max\left\{0,-ln\left(p\left(A|C\right)\right)-1\right\}]\end{array}$

Parameters on which the original FLoSS and the adapted FLoSS are based are shown and compared. The mathematical formula of the original FLoSS represents the sum of all deceased members, all living members of a sibship and a bonus for people still alive. The components are calculated from the negative ln function of a person's probability of reaching a special age in their birth-cohort C (same birthyear, gender and country of birth). A is the current age because only survival to age A is certain. In contrast, A* is the expected age at death (taken from life-table-calculations). $\left(p\left(A|C\right)\right)$ describes the probability that a random person in the same birth-cohort will live to age A. The ln-function guarantees a higher increase of the original FLoSS the higher the age of the individual. If the 1 is not subtracted the original FLoSS would be always positive, even for a person who died at a young age. Sebastiani et al. decided to subtract 1 to measure the exceptionality of a person's survival. The bonus $max\left\{0,-ln\left(p\left(A|C\right)\right)-1\right\}$ only increases the original FLoSS if the value is positive. The A on the bonus is the current age, not the expected age A*. The adapted FLoSS uses information of two generations (grand- and great-grandparents). In the first term ‘dead family members’ we investigate in deceased grand- and great-grandparents. Accordingly, in the second term we take living grand- and great-grandparents into account. In our birth-cohort the ‘living sibling bonus’ increases the score with exceptionally long-lived grand- and great-grandparents.

FLoSS: Family longevity selection score; KKS: KUNO-Kids Study.

We calculated the adapted FLoSS with all participants of the KUNO-Kids study who provided enough data. Families who gave answers for less than ten family-members were not considered. Additionally, we rejected implausible responses such as obviously incorrect annual data or conflicting answers.

The adapted FLoSS considers the following: The chance to reach a certain age in every birthyear was calculated from the cohort life tables from the German Federal Statistical Office [23]. We compared the age reached by the family members of our study subjects to the general German population of the same birth year. However, since the cohort life tables reached back only to the birthyear 1871, we had to impute data for the year 1870, for which we used the probability of living up to an individual age from the last year on record (1871) as estimates.

2.4. Application of the LRC score to a birth cohort

A longevity score alternative to FLoSS is the LRC score and we present its mathematical formula in Figure 2A. The LRC score weights the number of top 10% ancestors divided by the weighted total number of ancestors and shows the number of how many of the ancestors were long-lived (LRC of 0.5 means that 50% were long-lived) [6]. The LRC was developed on the basis of the evidence-based threshold at which longevity becomes heritable, namely by the top 10% survivors of their respective birth cohort [24] and LRC can be applied over several generations and is independent of the family-size [6]. One only needs to know whether a person's age is above or below the 90th percentile of their birth cohort. This is expressed in binary numbers (i.e., if a person reaches at least the 90th percentile of their birth cohort, they score 1, if they are below this threshold, they score 0). Additionally, the degree of relationship is considered by a relationship coefficient [6]. These data are usually available in a birth cohort setting.

Figure 2.

LRC score’s (A) mathematical formula, (B) its application in KKS with distribution of LRC values and (C) comparison of values and thresholds between adapted FLoSS and LRC scores in KKS. The LRC score describes the ratio of the number of long-lived individuals in a family (the top 10% of a birth cohort) to the total number of ancestors. In (A) the mathematical formula for LR is presented where I refers to the person who is included in the calculation (grandparents/great-grandparents)), i refers to the offspring for which the score is calculated (newborn). k is an index that refers to every ancestor (grandparents and great-grandparents of person i). p_k is the sex and birth year-specific survival percentile based on life tables of relative k. The relationship coefficients are used as weights w_k. For example, parents and children each share 50% of their DNA, grandparents and grandchildren 25% and great-grandparents and great-grandchildren 12.5% of their DNA. $l{\left(p_k\ge 0.9\right)}_i$ indicates whether relative k belongs to the top 10% survivors and ${\Sigma }_{k=1}^{N_i}{w}_k$ is the weighted total number of ancestors of the descendant i. In (B) the LRC is applied to the KKS and density of values is shown, where μ defines the mean, б the standard deviation.(C) shows a scatter plot of the adapted FLoSS against the LRC score in KKS. Thresholds for long-lived families are a value of ≥7 for adapted FLoSS and 0.2 for LRC.

FLoSS: Family longevity selection score; KKS: KUNO-Kids Study; LRC: Longevity Relatives Count.

Data taken from [19,24].

2.5. Statistics

Descriptive analysis was performed using proportions with frequencies for categorical data and means with standard deviations, median and interquartile range for metric data, respectively. Missing values were not imputed other than the cohort life tables before 1871 as reported above; the number of valid values is reported for all variables. Data were analysed using SPSS.24. We also calculated adapted FLoSS and LRC scores via SPSS.24.

3. Results

3.1. Study population & response rates at three different data time points

In the KKS, 3100 families participated as of December 2020. A total of 3072 (99.1%) of mothers took part in the interview within 48 h after birth (Figure 1A). As expected, the number of participants' responses decreased over time. After 4 weeks, 65.4 % of mothers and 61.1% of fathers participated in the next KUNO-Kids questionnaire survey. For the 1-year questionnaire, we received feedback from 44.2% of mothers and 43.3% of fathers, respectively.

3.2. Best time for data collection for familial longevity

As the quality of the score depends on the quality and quantity of information received about older generations and no previous experience existed how to best gather that information in a birth cohort setting, we explored the best timepoint for such a data collection. We questioned parents at different timepoints and different constellations (within 48 h after giving birth as well as 4 weeks and 1 year later). From all parents for whom data of all three timepoints were available, 50 were selected randomly to compare plausibility (e.g., digit transposition, unexpectedly large or small numbers) agreement between answers at different timepoints and completeness of the data at all three timepoints (Figure 1B). For that comparison in this subsample, the question on the grandparents' year of birth was assessed, as this was the question asked identically at all three time points. In our subsample, the questionnaire after birth yielded the best results with the most plausible as well as consistent data and the least missing answers for paternal and maternal lineage. Answers on maternal lineage were incomplete less often and completeness of paternal information was not increased by sending out questionnaires to fathers directly (4-weeks and 1-year questionnaires).

3.2. Adaptation of the FLoSS to a birth cohort setting

We calculated the adapted FLoSS for KKS applying some adaptations (Table 2): The main modification is that our index person is the newborn and we assess grandparents (third generation ancestor) and great-grandparents (fourth generation ancestor) without giving weight to siblings, as the index-person is not of old age. Therefore, in the first term of the formula ‘dead family members’ deceased grandparents and great-grandparents were included. The second part of the score ‘living family members’ also considers the living grandparents and great-grandparents. In the adapted FLoSS we applied the ‘living sibling bonus’ given by the original FLoSS for particularly old family members who are still alive, to grandparents and great-grandparents. Unnatural causes of death such as accidents, suicide and death in war affected 3.3% of the dead ancestors in our population. Of these 3.3%, a total of 13.5% (relating to 0.4% of the overall population) died under the age of 40 years and did not contribute to increase the adapted FLoSS as well as the 5.7% of relatives with missing data (0.19% of overall population). Furthermore, we excluding families with a migration background, defined as mother or father not born in Germany, from calculations [25].

Table 2.

Comparison between adapted FLoSS in KKS and original FLoSS in 3 different studies.

	KKS	FHS	LLFS	NECS	Refs.
Number of participants	551	766	1671	660
FLoSS ≥7 (%)	3.3	0.2	30	40
Mean μ	-0.15	0.00	3.92	5.37
Standard deviation б	2.33	1.00	3.22	3.36
Average age	72	74	83	91
Study description	Prospective birth cohort study in Eastern Bavaria, Germany	Systematic longitudinally followed examination of residents of the city of Framingham, Massachusetts and their offspring for heart diseases	US-Danish study of long-lived sibships and offspring	Sample of centenarians, their siblings and offspring in the Boston area, USA	[19,25]
Data for the FLoSS	Two generations (great-grandparents and grandparents of newborn) based on a birth cohort	At least one sibling born before 1925 and his siblings	Probands of at least 80 years and their siblings	Probands ≥100 years and their siblings	[19]

In the top graphs, s-FLoSS values of each study are plotted at the x-axis and density at the y-axis. Small boxes show the distribution of the standardized FLoSS (s-FLoSS = (FLoSS-M)/S; M = -0.24, S = 1.47), where M is the mean and S is the standard deviation of original FLoSS scores in the FHS cohort. μ defines the mean, б the standard deviation. The dashed lines show the standard normal density, the dotted lines the normal density with mean and standard deviation of the s-FLoSS. s-FLoSS graphs for FHS, LLFS and NECS. The table below the graphs summarizes and compares specific characteristics of KKS, FHS, LLFS and NECS in respect to (s)-FLoSS calculations.

FLoSS: Family longevity selection score; FHS: Framingham Heart Study; KKS: KUNO-Kids Study; LLFS: Long Life Family Study; NECS: New England Centenarian Study.

Data taken from [14,19].

3.3. Adapted FLoSS application in the KUNO-Kids study & comparison to other cohorts

After exclusion of families who did not provide enough plausible data, 551 families contributed to the calculations of the adapted FLoSS KKS. For this calculation, we used the data from the 1-year questionnaire, as we asked for the exact year of birth and death of all persons we needed for the adapted FLoSS-calculation in this questionnaire. This allowed for direct comparison to other studies. With a mean of -0.15, the adapted FLoSS KKS has a similar age structure as is to be expected in the normal population. The standard deviation of 2.33 shows a wide dispersion of the adapted FLoSS in KKS. In our cohort there are a few very short-lived and also a few extremely long-lived families, but most have a normal life expectation and an adapted FLoSS between -3 and +3 with a median of -0.48. With an average age of our grandparents and great-grandparents of 72 years, we identified 3.3% (18 out of 551) of families to have an adapted FLoSS KKS of ≥7, which is, as in the original FLoSS, considered to identify exceptional longevity [19].

Next, we compared the results of the adapted FLoSS in KKS with three cohorts of old individuals (Table 2), where an original FLoSS was calculated previously, namely the Framingham Heart Study (FHS), the Long Life Family Study (LLFS) and the New England Centenarian Study (NECS). To be able to compare all these studies, we used the s-FLoSS, which is the standardized FLoSS (s-FLoSS = (FLoSS-M)/S; M = -0.24, S = 1.47, where M is the mean and S is the standard deviation of original FLoSS in the FHS cohort) [19]. Results in KKS were closest to values of FHS, which, just like KKS, is a cohort consisting of people with a normal life expectancy and not of particularly long-lived individuals such as LLFS and NECS (KKS FLoSS mean -0.15, FHS-s-FLoSS mean 0.00, LLFS s-FLoSS mean 3.92, NECS s-FLoSS mean 5.37). Our standard deviation was 2.33 for the KKS, whereas the standard deviation is 1.00 for FHS, 3.22 for LLFS and 3.36 for NECS. Also, the s-FLoSS-value of 7 for exceptional longevity in KKS (3.3%) showed similar frequencies in FHS (0.2%) respectively, compared with 30% in LLFS and 40% in NECS,which are enriched for longevity.

Furthermore, we calculated the LRC score for our KKS cohort (Figure 2B) and plotted its results against the results of the adapted FLoSS in the same cohort (Figure 2C). When comparing the results of the adapted FLoSS and the LRC score, the same families tend to be long-lived, although there are deviations: Most families classified as long-lived by the adapted FLoSS also achieved higher LRC values (Supplementary Table S2). Thus, from an adapted FLoSS of -1 (average to rather short-lived families), almost only LRC values of 0 (that means no family member is long-lived) can be found (Figure 2B & C). A value of at least 7 in the adapted FLoSS is the threshold to be regarded as a family with longevity as it is in the original FLoSS. In our study population, 18 families reached this threshold. The LRC threshold to transmit longevity for at least two subsequent generations is at least 0.20, which means that 20 % of family members are long-lived (according to previous studies the top 10 % of their birth cohort are long-lived [24]). In our population, 13 families achieved this value. Of these families, six also have an adapted FLoSS >7 (Supplementary Table S2).

4. Discussion

In this study we showed that an adapted FLoSS can be calculated in a birth cohort and may be used in the future to study unbiased predictors of longevity in newborns prospectively. We found that the most pragmatic time point to collect data from ancestors to calculate an adapted FLoSS is directly at birth and that mothers are the best providers of all such data. When comparing the adapted FLoSS of our birth cohort with study populations where the original FLoSS was applied to much older study participants, the number of long-lived families identified in our cohort was similar to other general population studies and lower than in those enriched with elderly of exceptional life span. While these results suggest that an adapted FLoSS is suitable to study longevity in a birth cohort, further comparisons implied that the LRC score did not perform that well.

To study longevity markers in newborns may seem far-fetched at first glance and they have no direct consequences for the newborn yet. However, a major advantage of such an early approach is that all index subjects are alive and biomaterial can be collected prospectively over their life span, so that biomarkers and predictors such as epigenetic signatures can be assessed before lifetime events have occurred, which may alter these signatures and impact on longevity. Active smoking may be such a life event: It changes epigenetic signatures [26] and has a strong and negative impact on longevity [27]. To distinguish pre-existing epigenetic predictors of longevity from marks of unhealthy lifestyle, early analysis of biomaterial is necessary. In addition, the longitudinal setting with prospective data collection allows determining the timing of potential preventive measures in the individual.

Thus, our first pragmatic question was, if and when such information can be collected reliably in the setting of a birth cohort study, as we hypothesized that knowledge on older generations may not be easily retrievable from young parents. In our subsample-analysis of 50 datasets we found that the timepoint for data collection does not really matter. As quality and completeness of data was not better at the later survey times, it seems feasible to acquire robust longevity data necessary to calculate an adapted FLoSS right at birth, which also has the advantage that much less data is lost to dropout. Thus, we suggest that right after birth is the best and most convenient time to collect this data from mothers.

A specific challenge for our adapted FLoSS calculation was the necessity to use the German birth registry as a reference [23,25] as a change in territory and population occurred specifically in the German setting over the last century, with incomplete death-statistics for the last years of World War II and in addition, ethnicity is not considered in these statistics. Despite that, the genetic background of the German population has been a rather robust central European background [28]. Furthermore, differences in living conditions between East and West Germany until reunification in 1990 led to differences in the mortality and life expectancy of people living in the two parts of Germany but we hypothesize, that familial rather than environmental factors may determine exceptional familial longevity over generations, as studied in our cohort.

As we did exclude early deaths due to unnatural causes from our adapted FLoSS calculation, we may have underestimated longevity in these families/persons (as they would have lived longer to a natural death). While this can easily be included and modelled in the future, we tried to be as conservative as possible for the comparison with other populations and scores in this study. In contrast to original FLoSS studies, we did not study older siblings within one generation, but a direct line of several generations. The grandparents of newborns are usually quite young (mean age in our study 63 years), so the adapted FLoSS-value of a newborn is difficult to predict based on that generation alone, as longevity can only be safely concluded when a person's final age is reached, which means that the person has died. Thus, data of great-grandparents are of particular importance in a birth cohort setting, as they are older and have already died in more cases, which allows for a final calculation of FLoSS values. However, data concerning more distant generations are frequently no longer known to young family members and are missing more often than the data of grandparents. All that leads to a less accurate (but more conservative) FLoSS-value and a loss of informative families due to missing data. In an attempt not to inflate the inaccuracy of the adapted FLoSS by excessive estimations, we decided to exclude families with a total of more than two missing values for age in the four grandparents and eight great-grandparents. This leads to a rather small selection of families included in the analysis but deflates the risk of overestimation longevity in these families and results in a conservative outcome.

While the original FLoSS only detects long-lived families within sibships, the formula was suggested to be applicable to other age-groups and study-populations and even multigenerational relatives by the developers of the original FLoSS. When comparing the FLoSS with the Family Risk Score on the basis of three study-populations (NECS, FHS and LLFS), the original FLoSS represented a good basis for familial longevity-studies but also to discover family-risks of many different diseases [19]. An original FLoSS over two generations was also applied to the Utah Population Database [20]. In this study with 234,155 participants, offspring from an exceptional (mother, father or both have a FLoSS value ≥7) and an ordinary group (parents have both a FLoSS value <7) were compared and offspring indeed showed a higher possibility to reach older ages than those from the ordinary group. However, lifespan is more similar between siblings than between parents and children and when calculating the original FLoSS, 57,192 sibships had a mean value of 0.234 with a standard deviation of 2.9, which is very similar to the results of the adapted FLoSS in our KUNO-Kids cohort (0.15). Even though all these populations are American, they show similar original FLoSS values as the adapted FLoSS in our German based birth study. We reason that while average life expectancies of populations differ due to differences in health systems, nutrition and other external factors, familial longevity in genetically similar populations, such as the European descent here, seem to be rather similar.

Comparing values from the original FLoSS to values from adapted FLoSS in one study would be a perfect validation of the adapted score. However, these populations are rare [29], as they not only have to be collected over decades but would also need to include extensive families over generations. One such unique study over generations is based on the Danish longevity-enriched families [29], where it is shown that offspring of longevity-enriched families have a general lowering of disease risk, especially for mental and behavioural illnesses. Unfortunately, such health data are not available in the older generations in our study setting and we encourage to test the adapted FLoSS as suggested here in multigenerational studies which may already exist with necessary data and in future birth cohorts, as the necessary information can be collected pragmatically, as we showed here.

Alternatively to the FLoSS, Van den Berg et al. developed the LRC score, which includes three generations [6]. Applied to a historical sample from the 19th century, the LRC score suggested that longevity is only transmitted over two generations when at least 20% of all family members are long-lived. The more long-lived ancestors there are in a family, the greater the chance to live to a very old age compared with the average population and the higher the possibility that genetic factors play a role [6]. While the FLoSS may favour large families as younger family members can also erroneously increase the value of a family's FLoSS value, the LRC increases the likelihood of assigning familial longevity when a family only has few members [21].

In our study population, 18 families reached the adapted FLoSS-value of at least 7, which is the threshold value for a family with longevity and 13 families achieved the LRC threshold to transmit longevity for at least two subsequent generations of 0.20, while of these families, six have both. How this may lead to a strong underestimation of longevity in our setting with the LRC can be seen when taking a closer look into our highest-ranking adapted FLoSS family (Supplementary Table S1, first line): In generation 3 (great-grandparents of the newborn) there are two people aged 99 years, one person aged 91, 88 and 86 respectively. All these old individuals count with a positive value to the FLoSS. However, only the two people aged 99 contribute to the LRC. That indicates that LRC ignores some very old individuals if they do not reach the 90th percentile, although these may also aggravate familial longevity. As the LRC is a binary value, a person situated in the top 10 % of their birth cohort always contributes to the score with 1, if not in the top 10 % the person contributes not to the score at all. On the other hand, relatives who were short-lived do not lose weight at all with the LRC. While the LRC does not take into account if a very old person is still alive (which is irrelevant in historical pedigrees), the original and adapted FLoSS gives these particularly old, still living people extra weight, which is the case in modern birth cohorts. Thus, we suggest that the adapted FLoSS is more suitable for capturing longevity in current birth cohorts than the LRC score.

5. Conclusion

In conclusion, the adapted FLoSS seems to be a good and pragmatic tool, which seems to be superior to the traditional LRC, to identify families with exceptional lifespan in an active birth cohort. The data to calculate adapted FLoSS can be collected directly after birth and results were comparable to studies centered on old index persons. Based on data limited to our birth cohort we suggest that an adapted FLoSS can be applied as a meaningful outcome to study unbiased epigenetic predictors for longevity in the future, potentially opening the door for personalized interventions in the future [30].

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/17501911.2024.2370760

Author contributions

MK did the study design. JCP, MM, SB, CA and AK collected the data. JCP, SB and VDG performed the statistical analysis and data interpretation. MK, JCP and VDG wrote the manuscript.

Financial disclosure

The authors have no financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Competing interests disclosure

The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Writing disclosure

No writing assistance was utilized in the production of this manuscript.

Ethical conduct of research

All study participants gave their written informed consent and all study procedures were approved by the medical ethics committee of the University of Regensburg (file number: 14-101-0347 and 19-1646-101).

Data availability statement

Deidentified participant data which were analysed for this manuscript can be obtained upon reasonable request from the corresponding author.